2 * Copyright 2010 Christoph Bumiller
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
17 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
20 * OTHER DEALINGS IN THE SOFTWARE.
24 #include <nouveau_drm.h>
25 #include <nvif/class.h>
26 #include "util/u_format.h"
27 #include "util/u_format_s3tc.h"
28 #include "pipe/p_screen.h"
30 #include "vl/vl_decoder.h"
31 #include "vl/vl_video_buffer.h"
33 #include "nouveau_vp3_video.h"
35 #include "nvc0/nvc0_context.h"
36 #include "nvc0/nvc0_screen.h"
38 #include "nvc0/mme/com9097.mme.h"
39 #include "nvc0/mme/com90c0.mme.h"
42 nvc0_screen_is_format_supported(struct pipe_screen
*pscreen
,
43 enum pipe_format format
,
44 enum pipe_texture_target target
,
45 unsigned sample_count
,
48 const struct util_format_description
*desc
= util_format_description(format
);
52 if (!(0x117 & (1 << sample_count
))) /* 0, 1, 2, 4 or 8 */
55 if (!util_format_is_supported(format
, bindings
))
58 if ((bindings
& PIPE_BIND_SAMPLER_VIEW
) && (target
!= PIPE_BUFFER
))
59 if (util_format_get_blocksizebits(format
) == 3 * 32)
62 if (bindings
& PIPE_BIND_LINEAR
)
63 if (util_format_is_depth_or_stencil(format
) ||
64 (target
!= PIPE_TEXTURE_1D
&&
65 target
!= PIPE_TEXTURE_2D
&&
66 target
!= PIPE_TEXTURE_RECT
) ||
70 /* Restrict ETC2 and ASTC formats here. These are only supported on GK20A.
72 if ((desc
->layout
== UTIL_FORMAT_LAYOUT_ETC
||
73 desc
->layout
== UTIL_FORMAT_LAYOUT_ASTC
) &&
74 /* The claim is that this should work on GM107 but it doesn't. Need to
75 * test further and figure out if it's a nouveau issue or a HW one.
76 nouveau_screen(pscreen)->class_3d < GM107_3D_CLASS &&
78 nouveau_screen(pscreen
)->class_3d
!= NVEA_3D_CLASS
)
81 /* transfers & shared are always supported */
82 bindings
&= ~(PIPE_BIND_TRANSFER_READ
|
83 PIPE_BIND_TRANSFER_WRITE
|
87 return (( nvc0_format_table
[format
].usage
|
88 nvc0_vertex_format
[format
].usage
) & bindings
) == bindings
;
92 nvc0_screen_get_param(struct pipe_screen
*pscreen
, enum pipe_cap param
)
94 const uint16_t class_3d
= nouveau_screen(pscreen
)->class_3d
;
95 struct nouveau_device
*dev
= nouveau_screen(pscreen
)->device
;
98 /* non-boolean caps */
99 case PIPE_CAP_MAX_TEXTURE_2D_LEVELS
:
100 case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS
:
102 case PIPE_CAP_MAX_TEXTURE_3D_LEVELS
:
103 return (class_3d
>= NVE4_3D_CLASS
) ? 13 : 12;
104 case PIPE_CAP_MAX_TEXTURE_ARRAY_LAYERS
:
106 case PIPE_CAP_MIN_TEXEL_OFFSET
:
108 case PIPE_CAP_MAX_TEXEL_OFFSET
:
110 case PIPE_CAP_MIN_TEXTURE_GATHER_OFFSET
:
112 case PIPE_CAP_MAX_TEXTURE_GATHER_OFFSET
:
114 case PIPE_CAP_MAX_TEXTURE_BUFFER_SIZE
:
115 return 128 * 1024 * 1024;
116 case PIPE_CAP_GLSL_FEATURE_LEVEL
:
118 case PIPE_CAP_MAX_RENDER_TARGETS
:
120 case PIPE_CAP_MAX_DUAL_SOURCE_RENDER_TARGETS
:
122 case PIPE_CAP_MAX_STREAM_OUTPUT_BUFFERS
:
124 case PIPE_CAP_MAX_STREAM_OUTPUT_SEPARATE_COMPONENTS
:
125 case PIPE_CAP_MAX_STREAM_OUTPUT_INTERLEAVED_COMPONENTS
:
127 case PIPE_CAP_MAX_GEOMETRY_OUTPUT_VERTICES
:
128 case PIPE_CAP_MAX_GEOMETRY_TOTAL_OUTPUT_COMPONENTS
:
130 case PIPE_CAP_MAX_VERTEX_STREAMS
:
132 case PIPE_CAP_MAX_VERTEX_ATTRIB_STRIDE
:
134 case PIPE_CAP_CONSTANT_BUFFER_OFFSET_ALIGNMENT
:
136 case PIPE_CAP_TEXTURE_BUFFER_OFFSET_ALIGNMENT
:
137 return 16; /* 256 for binding as RT, but that's not possible in GL */
138 case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT
:
140 case PIPE_CAP_MIN_MAP_BUFFER_ALIGNMENT
:
141 return NOUVEAU_MIN_BUFFER_MAP_ALIGN
;
142 case PIPE_CAP_MAX_VIEWPORTS
:
143 return NVC0_MAX_VIEWPORTS
;
144 case PIPE_CAP_MAX_TEXTURE_GATHER_COMPONENTS
:
146 case PIPE_CAP_TEXTURE_BORDER_COLOR_QUIRK
:
147 return PIPE_QUIRK_TEXTURE_BORDER_COLOR_SWIZZLE_NV50
;
148 case PIPE_CAP_ENDIANNESS
:
149 return PIPE_ENDIAN_LITTLE
;
150 case PIPE_CAP_MAX_SHADER_PATCH_VARYINGS
:
154 case PIPE_CAP_TEXTURE_MIRROR_CLAMP
:
155 case PIPE_CAP_TEXTURE_SWIZZLE
:
156 case PIPE_CAP_TEXTURE_SHADOW_MAP
:
157 case PIPE_CAP_NPOT_TEXTURES
:
158 case PIPE_CAP_MIXED_FRAMEBUFFER_SIZES
:
159 case PIPE_CAP_ANISOTROPIC_FILTER
:
160 case PIPE_CAP_SEAMLESS_CUBE_MAP
:
161 case PIPE_CAP_CUBE_MAP_ARRAY
:
162 case PIPE_CAP_TEXTURE_BUFFER_OBJECTS
:
163 case PIPE_CAP_TEXTURE_MULTISAMPLE
:
164 case PIPE_CAP_TWO_SIDED_STENCIL
:
165 case PIPE_CAP_DEPTH_CLIP_DISABLE
:
166 case PIPE_CAP_POINT_SPRITE
:
167 case PIPE_CAP_TGSI_TEXCOORD
:
169 case PIPE_CAP_FRAGMENT_COLOR_CLAMPED
:
170 case PIPE_CAP_VERTEX_COLOR_UNCLAMPED
:
171 case PIPE_CAP_VERTEX_COLOR_CLAMPED
:
172 case PIPE_CAP_QUERY_TIMESTAMP
:
173 case PIPE_CAP_QUERY_TIME_ELAPSED
:
174 case PIPE_CAP_OCCLUSION_QUERY
:
175 case PIPE_CAP_STREAM_OUTPUT_PAUSE_RESUME
:
176 case PIPE_CAP_QUERY_PIPELINE_STATISTICS
:
177 case PIPE_CAP_BLEND_EQUATION_SEPARATE
:
178 case PIPE_CAP_INDEP_BLEND_ENABLE
:
179 case PIPE_CAP_INDEP_BLEND_FUNC
:
180 case PIPE_CAP_TGSI_FS_COORD_ORIGIN_UPPER_LEFT
:
181 case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_HALF_INTEGER
:
182 case PIPE_CAP_PRIMITIVE_RESTART
:
183 case PIPE_CAP_TGSI_INSTANCEID
:
184 case PIPE_CAP_VERTEX_ELEMENT_INSTANCE_DIVISOR
:
185 case PIPE_CAP_MIXED_COLORBUFFER_FORMATS
:
186 case PIPE_CAP_CONDITIONAL_RENDER
:
187 case PIPE_CAP_TEXTURE_BARRIER
:
188 case PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION
:
189 case PIPE_CAP_START_INSTANCE
:
190 case PIPE_CAP_BUFFER_MAP_PERSISTENT_COHERENT
:
191 case PIPE_CAP_DRAW_INDIRECT
:
192 case PIPE_CAP_USER_CONSTANT_BUFFERS
:
193 case PIPE_CAP_USER_INDEX_BUFFERS
:
194 case PIPE_CAP_USER_VERTEX_BUFFERS
:
195 case PIPE_CAP_TEXTURE_QUERY_LOD
:
196 case PIPE_CAP_SAMPLE_SHADING
:
197 case PIPE_CAP_TEXTURE_GATHER_OFFSETS
:
198 case PIPE_CAP_TEXTURE_GATHER_SM5
:
199 case PIPE_CAP_TGSI_FS_FINE_DERIVATIVE
:
200 case PIPE_CAP_CONDITIONAL_RENDER_INVERTED
:
201 case PIPE_CAP_SAMPLER_VIEW_TARGET
:
202 case PIPE_CAP_CLIP_HALFZ
:
203 case PIPE_CAP_POLYGON_OFFSET_CLAMP
:
204 case PIPE_CAP_MULTISAMPLE_Z_RESOLVE
:
205 case PIPE_CAP_TEXTURE_FLOAT_LINEAR
:
206 case PIPE_CAP_TEXTURE_HALF_FLOAT_LINEAR
:
207 case PIPE_CAP_DEPTH_BOUNDS_TEST
:
208 case PIPE_CAP_TGSI_TXQS
:
209 case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS
:
210 case PIPE_CAP_FORCE_PERSAMPLE_INTERP
:
211 case PIPE_CAP_SHAREABLE_SHADERS
:
212 case PIPE_CAP_CLEAR_TEXTURE
:
213 case PIPE_CAP_DRAW_PARAMETERS
:
214 case PIPE_CAP_TGSI_PACK_HALF_FLOAT
:
215 case PIPE_CAP_MULTI_DRAW_INDIRECT
:
216 case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS
:
217 case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL
:
218 case PIPE_CAP_QUERY_BUFFER_OBJECT
:
220 case PIPE_CAP_SEAMLESS_CUBE_MAP_PER_TEXTURE
:
221 return (class_3d
>= NVE4_3D_CLASS
) ? 1 : 0;
222 case PIPE_CAP_COMPUTE
:
223 if (debug_get_bool_option("NVF0_COMPUTE", false))
225 return (class_3d
<= NVE4_3D_CLASS
) ? 1 : 0;
226 case PIPE_CAP_PREFER_BLIT_BASED_TEXTURE_TRANSFER
:
227 return nouveau_screen(pscreen
)->vram_domain
& NOUVEAU_BO_VRAM
? 1 : 0;
229 /* unsupported caps */
230 case PIPE_CAP_TGSI_FS_COORD_ORIGIN_LOWER_LEFT
:
231 case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER
:
232 case PIPE_CAP_SHADER_STENCIL_EXPORT
:
233 case PIPE_CAP_TGSI_CAN_COMPACT_CONSTANTS
:
234 case PIPE_CAP_VERTEX_BUFFER_OFFSET_4BYTE_ALIGNED_ONLY
:
235 case PIPE_CAP_VERTEX_BUFFER_STRIDE_4BYTE_ALIGNED_ONLY
:
236 case PIPE_CAP_VERTEX_ELEMENT_SRC_OFFSET_4BYTE_ALIGNED_ONLY
:
237 case PIPE_CAP_TGSI_VS_LAYER_VIEWPORT
:
238 case PIPE_CAP_FAKE_SW_MSAA
:
239 case PIPE_CAP_TGSI_VS_WINDOW_SPACE_POSITION
:
240 case PIPE_CAP_VERTEXID_NOBASE
:
241 case PIPE_CAP_RESOURCE_FROM_USER_MEMORY
:
242 case PIPE_CAP_DEVICE_RESET_STATUS_QUERY
:
243 case PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL
:
244 case PIPE_CAP_INVALIDATE_BUFFER
:
245 case PIPE_CAP_GENERATE_MIPMAP
:
246 case PIPE_CAP_STRING_MARKER
:
247 case PIPE_CAP_BUFFER_SAMPLER_VIEW_RGBA_ONLY
:
248 case PIPE_CAP_SURFACE_REINTERPRET_BLOCKS
:
249 case PIPE_CAP_QUERY_MEMORY_INFO
:
250 case PIPE_CAP_PCI_GROUP
:
251 case PIPE_CAP_PCI_BUS
:
252 case PIPE_CAP_PCI_DEVICE
:
253 case PIPE_CAP_PCI_FUNCTION
:
256 case PIPE_CAP_VENDOR_ID
:
258 case PIPE_CAP_DEVICE_ID
: {
260 if (nouveau_getparam(dev
, NOUVEAU_GETPARAM_PCI_DEVICE
, &device_id
)) {
261 NOUVEAU_ERR("NOUVEAU_GETPARAM_PCI_DEVICE failed.\n");
266 case PIPE_CAP_ACCELERATED
:
268 case PIPE_CAP_VIDEO_MEMORY
:
269 return dev
->vram_size
>> 20;
274 NOUVEAU_ERR("unknown PIPE_CAP %d\n", param
);
279 nvc0_screen_get_shader_param(struct pipe_screen
*pscreen
, unsigned shader
,
280 enum pipe_shader_cap param
)
282 const uint16_t class_3d
= nouveau_screen(pscreen
)->class_3d
;
285 case PIPE_SHADER_VERTEX
:
286 case PIPE_SHADER_GEOMETRY
:
287 case PIPE_SHADER_FRAGMENT
:
289 case PIPE_SHADER_TESS_CTRL
:
290 case PIPE_SHADER_TESS_EVAL
:
291 if (class_3d
>= GM107_3D_CLASS
)
294 case PIPE_SHADER_COMPUTE
:
295 if (!debug_get_bool_option("NVF0_COMPUTE", false))
296 if (class_3d
> NVE4_3D_CLASS
)
304 case PIPE_SHADER_CAP_PREFERRED_IR
:
305 return PIPE_SHADER_IR_TGSI
;
306 case PIPE_SHADER_CAP_SUPPORTED_IRS
:
307 if (class_3d
== NVF0_3D_CLASS
&&
308 !debug_get_bool_option("NVF0_COMPUTE", false))
310 return 1 << PIPE_SHADER_IR_TGSI
;
311 case PIPE_SHADER_CAP_MAX_INSTRUCTIONS
:
312 case PIPE_SHADER_CAP_MAX_ALU_INSTRUCTIONS
:
313 case PIPE_SHADER_CAP_MAX_TEX_INSTRUCTIONS
:
314 case PIPE_SHADER_CAP_MAX_TEX_INDIRECTIONS
:
316 case PIPE_SHADER_CAP_MAX_CONTROL_FLOW_DEPTH
:
318 case PIPE_SHADER_CAP_MAX_INPUTS
:
319 if (shader
== PIPE_SHADER_VERTEX
)
321 /* NOTE: These only count our slots for GENERIC varyings.
322 * The address space may be larger, but the actual hard limit seems to be
323 * less than what the address space layout permits, so don't add TEXCOORD,
326 if (shader
== PIPE_SHADER_FRAGMENT
)
328 /* Actually this counts CLIPVERTEX, which occupies the last generic slot,
329 * and excludes 0x60 per-patch inputs.
332 case PIPE_SHADER_CAP_MAX_OUTPUTS
:
334 case PIPE_SHADER_CAP_MAX_CONST_BUFFER_SIZE
:
336 case PIPE_SHADER_CAP_MAX_CONST_BUFFERS
:
337 return NVC0_MAX_PIPE_CONSTBUFS
;
338 case PIPE_SHADER_CAP_INDIRECT_OUTPUT_ADDR
:
339 return shader
!= PIPE_SHADER_FRAGMENT
;
340 case PIPE_SHADER_CAP_INDIRECT_INPUT_ADDR
:
341 return shader
!= PIPE_SHADER_FRAGMENT
|| class_3d
< GM107_3D_CLASS
;
342 case PIPE_SHADER_CAP_INDIRECT_TEMP_ADDR
:
343 case PIPE_SHADER_CAP_INDIRECT_CONST_ADDR
:
345 case PIPE_SHADER_CAP_MAX_PREDS
:
347 case PIPE_SHADER_CAP_MAX_TEMPS
:
348 return NVC0_CAP_MAX_PROGRAM_TEMPS
;
349 case PIPE_SHADER_CAP_TGSI_CONT_SUPPORTED
:
351 case PIPE_SHADER_CAP_TGSI_SQRT_SUPPORTED
:
353 case PIPE_SHADER_CAP_SUBROUTINES
:
355 case PIPE_SHADER_CAP_INTEGERS
:
357 case PIPE_SHADER_CAP_DOUBLES
:
359 case PIPE_SHADER_CAP_TGSI_DROUND_SUPPORTED
:
361 case PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED
:
363 case PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED
:
364 case PIPE_SHADER_CAP_TGSI_ANY_INOUT_DECL_RANGE
:
366 case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS
:
367 return NVC0_MAX_BUFFERS
;
368 case PIPE_SHADER_CAP_MAX_TEXTURE_SAMPLERS
:
369 return 16; /* would be 32 in linked (OpenGL-style) mode */
370 case PIPE_SHADER_CAP_MAX_SAMPLER_VIEWS
:
371 return 16; /* XXX not sure if more are really safe */
372 case PIPE_SHADER_CAP_MAX_UNROLL_ITERATIONS_HINT
:
374 case PIPE_SHADER_CAP_MAX_SHADER_IMAGES
:
377 NOUVEAU_ERR("unknown PIPE_SHADER_CAP %d\n", param
);
383 nvc0_screen_get_paramf(struct pipe_screen
*pscreen
, enum pipe_capf param
)
386 case PIPE_CAPF_MAX_LINE_WIDTH
:
387 case PIPE_CAPF_MAX_LINE_WIDTH_AA
:
389 case PIPE_CAPF_MAX_POINT_WIDTH
:
391 case PIPE_CAPF_MAX_POINT_WIDTH_AA
:
393 case PIPE_CAPF_MAX_TEXTURE_ANISOTROPY
:
395 case PIPE_CAPF_MAX_TEXTURE_LOD_BIAS
:
397 case PIPE_CAPF_GUARD_BAND_LEFT
:
398 case PIPE_CAPF_GUARD_BAND_TOP
:
400 case PIPE_CAPF_GUARD_BAND_RIGHT
:
401 case PIPE_CAPF_GUARD_BAND_BOTTOM
:
402 return 0.0f
; /* that or infinity */
405 NOUVEAU_ERR("unknown PIPE_CAPF %d\n", param
);
410 nvc0_screen_get_compute_param(struct pipe_screen
*pscreen
,
411 enum pipe_shader_ir ir_type
,
412 enum pipe_compute_cap param
, void *data
)
414 struct nvc0_screen
*screen
= nvc0_screen(pscreen
);
415 const uint16_t obj_class
= screen
->compute
->oclass
;
417 #define RET(x) do { \
419 memcpy(data, x, sizeof(x)); \
424 case PIPE_COMPUTE_CAP_GRID_DIMENSION
:
425 RET((uint64_t []) { 3 });
426 case PIPE_COMPUTE_CAP_MAX_GRID_SIZE
:
427 if (obj_class
>= NVE4_COMPUTE_CLASS
) {
428 RET(((uint64_t []) { 0x7fffffff, 65535, 65535 }));
430 RET(((uint64_t []) { 65535, 65535, 65535 }));
432 case PIPE_COMPUTE_CAP_MAX_BLOCK_SIZE
:
433 RET(((uint64_t []) { 1024, 1024, 64 }));
434 case PIPE_COMPUTE_CAP_MAX_THREADS_PER_BLOCK
:
435 RET((uint64_t []) { 1024 });
436 case PIPE_COMPUTE_CAP_MAX_GLOBAL_SIZE
: /* g[] */
437 RET((uint64_t []) { 1ULL << 40 });
438 case PIPE_COMPUTE_CAP_MAX_LOCAL_SIZE
: /* s[] */
439 RET((uint64_t []) { 48 << 10 });
440 case PIPE_COMPUTE_CAP_MAX_PRIVATE_SIZE
: /* l[] */
441 RET((uint64_t []) { 512 << 10 });
442 case PIPE_COMPUTE_CAP_MAX_INPUT_SIZE
: /* c[], arbitrary limit */
443 RET((uint64_t []) { 4096 });
444 case PIPE_COMPUTE_CAP_SUBGROUP_SIZE
:
445 RET((uint32_t []) { 32 });
446 case PIPE_COMPUTE_CAP_MAX_MEM_ALLOC_SIZE
:
447 RET((uint64_t []) { 1ULL << 40 });
448 case PIPE_COMPUTE_CAP_IMAGES_SUPPORTED
:
449 RET((uint32_t []) { 0 });
450 case PIPE_COMPUTE_CAP_MAX_COMPUTE_UNITS
:
451 RET((uint32_t []) { screen
->mp_count_compute
});
452 case PIPE_COMPUTE_CAP_MAX_CLOCK_FREQUENCY
:
453 RET((uint32_t []) { 512 }); /* FIXME: arbitrary limit */
462 nvc0_screen_destroy(struct pipe_screen
*pscreen
)
464 struct nvc0_screen
*screen
= nvc0_screen(pscreen
);
466 if (!nouveau_drm_screen_unref(&screen
->base
))
469 if (screen
->base
.fence
.current
) {
470 struct nouveau_fence
*current
= NULL
;
472 /* nouveau_fence_wait will create a new current fence, so wait on the
473 * _current_ one, and remove both.
475 nouveau_fence_ref(screen
->base
.fence
.current
, ¤t
);
476 nouveau_fence_wait(current
, NULL
);
477 nouveau_fence_ref(NULL
, ¤t
);
478 nouveau_fence_ref(NULL
, &screen
->base
.fence
.current
);
480 if (screen
->base
.pushbuf
)
481 screen
->base
.pushbuf
->user_priv
= NULL
;
484 nvc0_blitter_destroy(screen
);
485 if (screen
->pm
.prog
) {
486 screen
->pm
.prog
->code
= NULL
; /* hardcoded, don't FREE */
487 nvc0_program_destroy(NULL
, screen
->pm
.prog
);
488 FREE(screen
->pm
.prog
);
491 nouveau_bo_ref(NULL
, &screen
->text
);
492 nouveau_bo_ref(NULL
, &screen
->uniform_bo
);
493 nouveau_bo_ref(NULL
, &screen
->tls
);
494 nouveau_bo_ref(NULL
, &screen
->txc
);
495 nouveau_bo_ref(NULL
, &screen
->fence
.bo
);
496 nouveau_bo_ref(NULL
, &screen
->poly_cache
);
497 nouveau_bo_ref(NULL
, &screen
->parm
);
499 nouveau_heap_destroy(&screen
->lib_code
);
500 nouveau_heap_destroy(&screen
->text_heap
);
502 FREE(screen
->tic
.entries
);
504 nouveau_object_del(&screen
->eng3d
);
505 nouveau_object_del(&screen
->eng2d
);
506 nouveau_object_del(&screen
->m2mf
);
507 nouveau_object_del(&screen
->compute
);
508 nouveau_object_del(&screen
->nvsw
);
510 nouveau_screen_fini(&screen
->base
);
516 nvc0_graph_set_macro(struct nvc0_screen
*screen
, uint32_t m
, unsigned pos
,
517 unsigned size
, const uint32_t *data
)
519 struct nouveau_pushbuf
*push
= screen
->base
.pushbuf
;
523 assert((pos
+ size
) <= 0x800);
525 BEGIN_NVC0(push
, SUBC_3D(NVC0_GRAPH_MACRO_ID
), 2);
526 PUSH_DATA (push
, (m
- 0x3800) / 8);
527 PUSH_DATA (push
, pos
);
528 BEGIN_1IC0(push
, SUBC_3D(NVC0_GRAPH_MACRO_UPLOAD_POS
), size
+ 1);
529 PUSH_DATA (push
, pos
);
530 PUSH_DATAp(push
, data
, size
);
536 nvc0_magic_3d_init(struct nouveau_pushbuf
*push
, uint16_t obj_class
)
538 BEGIN_NVC0(push
, SUBC_3D(0x10cc), 1);
539 PUSH_DATA (push
, 0xff);
540 BEGIN_NVC0(push
, SUBC_3D(0x10e0), 2);
541 PUSH_DATA (push
, 0xff);
542 PUSH_DATA (push
, 0xff);
543 BEGIN_NVC0(push
, SUBC_3D(0x10ec), 2);
544 PUSH_DATA (push
, 0xff);
545 PUSH_DATA (push
, 0xff);
546 BEGIN_NVC0(push
, SUBC_3D(0x074c), 1);
547 PUSH_DATA (push
, 0x3f);
549 BEGIN_NVC0(push
, SUBC_3D(0x16a8), 1);
550 PUSH_DATA (push
, (3 << 16) | 3);
551 BEGIN_NVC0(push
, SUBC_3D(0x1794), 1);
552 PUSH_DATA (push
, (2 << 16) | 2);
554 if (obj_class
< GM107_3D_CLASS
) {
555 BEGIN_NVC0(push
, SUBC_3D(0x12ac), 1);
558 BEGIN_NVC0(push
, SUBC_3D(0x0218), 1);
559 PUSH_DATA (push
, 0x10);
560 BEGIN_NVC0(push
, SUBC_3D(0x10fc), 1);
561 PUSH_DATA (push
, 0x10);
562 BEGIN_NVC0(push
, SUBC_3D(0x1290), 1);
563 PUSH_DATA (push
, 0x10);
564 BEGIN_NVC0(push
, SUBC_3D(0x12d8), 2);
565 PUSH_DATA (push
, 0x10);
566 PUSH_DATA (push
, 0x10);
567 BEGIN_NVC0(push
, SUBC_3D(0x1140), 1);
568 PUSH_DATA (push
, 0x10);
569 BEGIN_NVC0(push
, SUBC_3D(0x1610), 1);
570 PUSH_DATA (push
, 0xe);
572 BEGIN_NVC0(push
, NVC0_3D(VERTEX_ID_GEN_MODE
), 1);
573 PUSH_DATA (push
, NVC0_3D_VERTEX_ID_GEN_MODE_DRAW_ARRAYS_ADD_START
);
574 BEGIN_NVC0(push
, SUBC_3D(0x030c), 1);
576 BEGIN_NVC0(push
, SUBC_3D(0x0300), 1);
579 BEGIN_NVC0(push
, SUBC_3D(0x02d0), 1);
580 PUSH_DATA (push
, 0x3fffff);
581 BEGIN_NVC0(push
, SUBC_3D(0x0fdc), 1);
583 BEGIN_NVC0(push
, SUBC_3D(0x19c0), 1);
586 if (obj_class
< GM107_3D_CLASS
) {
587 BEGIN_NVC0(push
, SUBC_3D(0x075c), 1);
590 if (obj_class
>= NVE4_3D_CLASS
) {
591 BEGIN_NVC0(push
, SUBC_3D(0x07fc), 1);
596 /* TODO: find out what software methods 0x1528, 0x1280 and (on nve4) 0x02dc
597 * are supposed to do */
601 nvc0_screen_fence_emit(struct pipe_screen
*pscreen
, u32
*sequence
)
603 struct nvc0_screen
*screen
= nvc0_screen(pscreen
);
604 struct nouveau_pushbuf
*push
= screen
->base
.pushbuf
;
606 /* we need to do it after possible flush in MARK_RING */
607 *sequence
= ++screen
->base
.fence
.sequence
;
609 assert(PUSH_AVAIL(push
) + push
->rsvd_kick
>= 5);
610 PUSH_DATA (push
, NVC0_FIFO_PKHDR_SQ(NVC0_3D(QUERY_ADDRESS_HIGH
), 4));
611 PUSH_DATAh(push
, screen
->fence
.bo
->offset
);
612 PUSH_DATA (push
, screen
->fence
.bo
->offset
);
613 PUSH_DATA (push
, *sequence
);
614 PUSH_DATA (push
, NVC0_3D_QUERY_GET_FENCE
| NVC0_3D_QUERY_GET_SHORT
|
615 (0xf << NVC0_3D_QUERY_GET_UNIT__SHIFT
));
619 nvc0_screen_fence_update(struct pipe_screen
*pscreen
)
621 struct nvc0_screen
*screen
= nvc0_screen(pscreen
);
622 return screen
->fence
.map
[0];
626 nvc0_screen_init_compute(struct nvc0_screen
*screen
)
628 screen
->base
.base
.get_compute_param
= nvc0_screen_get_compute_param
;
630 switch (screen
->base
.device
->chipset
& ~0xf) {
633 return nvc0_screen_compute_setup(screen
, screen
->base
.pushbuf
);
635 return nve4_screen_compute_setup(screen
, screen
->base
.pushbuf
);
639 if (debug_get_bool_option("NVF0_COMPUTE", false))
640 return nve4_screen_compute_setup(screen
, screen
->base
.pushbuf
);
649 nvc0_screen_resize_tls_area(struct nvc0_screen
*screen
,
650 uint32_t lpos
, uint32_t lneg
, uint32_t cstack
)
652 struct nouveau_bo
*bo
= NULL
;
654 uint64_t size
= (lpos
+ lneg
) * 32 + cstack
;
656 if (size
>= (1 << 20)) {
657 NOUVEAU_ERR("requested TLS size too large: 0x%"PRIx64
"\n", size
);
661 size
*= (screen
->base
.device
->chipset
>= 0xe0) ? 64 : 48; /* max warps */
662 size
= align(size
, 0x8000);
663 size
*= screen
->mp_count
;
665 size
= align(size
, 1 << 17);
667 ret
= nouveau_bo_new(screen
->base
.device
, NV_VRAM_DOMAIN(&screen
->base
), 1 << 17, size
,
670 NOUVEAU_ERR("failed to allocate TLS area, size: 0x%"PRIx64
"\n", size
);
673 nouveau_bo_ref(NULL
, &screen
->tls
);
678 #define FAIL_SCREEN_INIT(str, err) \
680 NOUVEAU_ERR(str, err); \
684 struct nouveau_screen
*
685 nvc0_screen_create(struct nouveau_device
*dev
)
687 struct nvc0_screen
*screen
;
688 struct pipe_screen
*pscreen
;
689 struct nouveau_object
*chan
;
690 struct nouveau_pushbuf
*push
;
697 switch (dev
->chipset
& ~0xf) {
710 screen
= CALLOC_STRUCT(nvc0_screen
);
713 pscreen
= &screen
->base
.base
;
714 pscreen
->destroy
= nvc0_screen_destroy
;
716 ret
= nouveau_screen_init(&screen
->base
, dev
);
718 nvc0_screen_destroy(pscreen
);
721 chan
= screen
->base
.channel
;
722 push
= screen
->base
.pushbuf
;
723 push
->user_priv
= screen
;
726 screen
->base
.vidmem_bindings
|= PIPE_BIND_CONSTANT_BUFFER
|
727 PIPE_BIND_SHADER_BUFFER
|
728 PIPE_BIND_VERTEX_BUFFER
| PIPE_BIND_INDEX_BUFFER
|
729 PIPE_BIND_COMMAND_ARGS_BUFFER
| PIPE_BIND_QUERY_BUFFER
;
730 screen
->base
.sysmem_bindings
|=
731 PIPE_BIND_VERTEX_BUFFER
| PIPE_BIND_INDEX_BUFFER
;
733 if (screen
->base
.vram_domain
& NOUVEAU_BO_GART
) {
734 screen
->base
.sysmem_bindings
|= screen
->base
.vidmem_bindings
;
735 screen
->base
.vidmem_bindings
= 0;
738 pscreen
->context_create
= nvc0_create
;
739 pscreen
->is_format_supported
= nvc0_screen_is_format_supported
;
740 pscreen
->get_param
= nvc0_screen_get_param
;
741 pscreen
->get_shader_param
= nvc0_screen_get_shader_param
;
742 pscreen
->get_paramf
= nvc0_screen_get_paramf
;
743 pscreen
->get_driver_query_info
= nvc0_screen_get_driver_query_info
;
744 pscreen
->get_driver_query_group_info
= nvc0_screen_get_driver_query_group_info
;
746 nvc0_screen_init_resource_functions(pscreen
);
748 screen
->base
.base
.get_video_param
= nouveau_vp3_screen_get_video_param
;
749 screen
->base
.base
.is_video_format_supported
= nouveau_vp3_screen_video_supported
;
751 flags
= NOUVEAU_BO_GART
| NOUVEAU_BO_MAP
;
752 if (screen
->base
.drm
->version
>= 0x01000202)
753 flags
|= NOUVEAU_BO_COHERENT
;
755 ret
= nouveau_bo_new(dev
, flags
, 0, 4096, NULL
, &screen
->fence
.bo
);
758 nouveau_bo_map(screen
->fence
.bo
, 0, NULL
);
759 screen
->fence
.map
= screen
->fence
.bo
->map
;
760 screen
->base
.fence
.emit
= nvc0_screen_fence_emit
;
761 screen
->base
.fence
.update
= nvc0_screen_fence_update
;
764 ret
= nouveau_object_new(chan
, (dev
->chipset
< 0xe0) ? 0x1f906e : 0x906e,
765 NVIF_CLASS_SW_GF100
, NULL
, 0, &screen
->nvsw
);
767 FAIL_SCREEN_INIT("Error creating SW object: %d\n", ret
);
769 BEGIN_NVC0(push
, SUBC_SW(NV01_SUBCHAN_OBJECT
), 1);
770 PUSH_DATA (push
, screen
->nvsw
->handle
);
772 switch (dev
->chipset
& ~0xf) {
777 obj_class
= NVF0_P2MF_CLASS
;
780 obj_class
= NVE4_P2MF_CLASS
;
783 obj_class
= NVC0_M2MF_CLASS
;
786 ret
= nouveau_object_new(chan
, 0xbeef323f, obj_class
, NULL
, 0,
789 FAIL_SCREEN_INIT("Error allocating PGRAPH context for M2MF: %d\n", ret
);
791 BEGIN_NVC0(push
, SUBC_M2MF(NV01_SUBCHAN_OBJECT
), 1);
792 PUSH_DATA (push
, screen
->m2mf
->oclass
);
793 if (screen
->m2mf
->oclass
== NVE4_P2MF_CLASS
) {
794 BEGIN_NVC0(push
, SUBC_COPY(NV01_SUBCHAN_OBJECT
), 1);
795 PUSH_DATA (push
, 0xa0b5);
798 ret
= nouveau_object_new(chan
, 0xbeef902d, NVC0_2D_CLASS
, NULL
, 0,
801 FAIL_SCREEN_INIT("Error allocating PGRAPH context for 2D: %d\n", ret
);
803 BEGIN_NVC0(push
, SUBC_2D(NV01_SUBCHAN_OBJECT
), 1);
804 PUSH_DATA (push
, screen
->eng2d
->oclass
);
805 BEGIN_NVC0(push
, SUBC_2D(NVC0_2D_SINGLE_GPC
), 1);
807 BEGIN_NVC0(push
, NVC0_2D(OPERATION
), 1);
808 PUSH_DATA (push
, NV50_2D_OPERATION_SRCCOPY
);
809 BEGIN_NVC0(push
, NVC0_2D(CLIP_ENABLE
), 1);
811 BEGIN_NVC0(push
, NVC0_2D(COLOR_KEY_ENABLE
), 1);
813 BEGIN_NVC0(push
, SUBC_2D(0x0884), 1);
814 PUSH_DATA (push
, 0x3f);
815 BEGIN_NVC0(push
, SUBC_2D(0x0888), 1);
817 BEGIN_NVC0(push
, NVC0_2D(COND_MODE
), 1);
818 PUSH_DATA (push
, NV50_2D_COND_MODE_ALWAYS
);
820 BEGIN_NVC0(push
, SUBC_2D(NVC0_GRAPH_NOTIFY_ADDRESS_HIGH
), 2);
821 PUSH_DATAh(push
, screen
->fence
.bo
->offset
+ 16);
822 PUSH_DATA (push
, screen
->fence
.bo
->offset
+ 16);
824 switch (dev
->chipset
& ~0xf) {
826 obj_class
= GM200_3D_CLASS
;
829 obj_class
= GM107_3D_CLASS
;
833 obj_class
= NVF0_3D_CLASS
;
836 switch (dev
->chipset
) {
838 obj_class
= NVEA_3D_CLASS
;
841 obj_class
= NVE4_3D_CLASS
;
846 obj_class
= NVC8_3D_CLASS
;
850 switch (dev
->chipset
) {
852 obj_class
= NVC8_3D_CLASS
;
855 obj_class
= NVC1_3D_CLASS
;
858 obj_class
= NVC0_3D_CLASS
;
863 ret
= nouveau_object_new(chan
, 0xbeef003d, obj_class
, NULL
, 0,
866 FAIL_SCREEN_INIT("Error allocating PGRAPH context for 3D: %d\n", ret
);
867 screen
->base
.class_3d
= obj_class
;
869 BEGIN_NVC0(push
, SUBC_3D(NV01_SUBCHAN_OBJECT
), 1);
870 PUSH_DATA (push
, screen
->eng3d
->oclass
);
872 BEGIN_NVC0(push
, NVC0_3D(COND_MODE
), 1);
873 PUSH_DATA (push
, NVC0_3D_COND_MODE_ALWAYS
);
875 if (debug_get_bool_option("NOUVEAU_SHADER_WATCHDOG", true)) {
876 /* kill shaders after about 1 second (at 100 MHz) */
877 BEGIN_NVC0(push
, NVC0_3D(WATCHDOG_TIMER
), 1);
878 PUSH_DATA (push
, 0x17);
881 IMMED_NVC0(push
, NVC0_3D(ZETA_COMP_ENABLE
),
882 screen
->base
.drm
->version
>= 0x01000101);
883 BEGIN_NVC0(push
, NVC0_3D(RT_COMP_ENABLE(0)), 8);
884 for (i
= 0; i
< 8; ++i
)
885 PUSH_DATA(push
, screen
->base
.drm
->version
>= 0x01000101);
887 BEGIN_NVC0(push
, NVC0_3D(RT_CONTROL
), 1);
890 BEGIN_NVC0(push
, NVC0_3D(CSAA_ENABLE
), 1);
892 BEGIN_NVC0(push
, NVC0_3D(MULTISAMPLE_ENABLE
), 1);
894 BEGIN_NVC0(push
, NVC0_3D(MULTISAMPLE_MODE
), 1);
895 PUSH_DATA (push
, NVC0_3D_MULTISAMPLE_MODE_MS1
);
896 BEGIN_NVC0(push
, NVC0_3D(MULTISAMPLE_CTRL
), 1);
898 BEGIN_NVC0(push
, NVC0_3D(LINE_WIDTH_SEPARATE
), 1);
900 BEGIN_NVC0(push
, NVC0_3D(PRIM_RESTART_WITH_DRAW_ARRAYS
), 1);
902 BEGIN_NVC0(push
, NVC0_3D(BLEND_SEPARATE_ALPHA
), 1);
904 BEGIN_NVC0(push
, NVC0_3D(BLEND_ENABLE_COMMON
), 1);
906 BEGIN_NVC0(push
, NVC0_3D(SHADE_MODEL
), 1);
907 PUSH_DATA (push
, NVC0_3D_SHADE_MODEL_SMOOTH
);
908 if (screen
->eng3d
->oclass
< NVE4_3D_CLASS
) {
909 IMMED_NVC0(push
, NVC0_3D(TEX_MISC
), 0);
911 BEGIN_NVC0(push
, NVE4_3D(TEX_CB_INDEX
), 1);
912 PUSH_DATA (push
, 15);
914 BEGIN_NVC0(push
, NVC0_3D(CALL_LIMIT_LOG
), 1);
915 PUSH_DATA (push
, 8); /* 128 */
916 BEGIN_NVC0(push
, NVC0_3D(ZCULL_STATCTRS_ENABLE
), 1);
918 if (screen
->eng3d
->oclass
>= NVC1_3D_CLASS
) {
919 BEGIN_NVC0(push
, NVC0_3D(CACHE_SPLIT
), 1);
920 PUSH_DATA (push
, NVC0_3D_CACHE_SPLIT_48K_SHARED_16K_L1
);
923 nvc0_magic_3d_init(push
, screen
->eng3d
->oclass
);
925 ret
= nouveau_bo_new(dev
, NV_VRAM_DOMAIN(&screen
->base
), 1 << 17, 1 << 20, NULL
,
930 /* XXX: getting a page fault at the end of the code buffer every few
931 * launches, don't use the last 256 bytes to work around them - prefetch ?
933 nouveau_heap_init(&screen
->text_heap
, 0, (1 << 20) - 0x100);
935 ret
= nouveau_bo_new(dev
, NV_VRAM_DOMAIN(&screen
->base
), 1 << 12, 7 << 16, NULL
,
936 &screen
->uniform_bo
);
940 PUSH_REFN (push
, screen
->uniform_bo
, NV_VRAM_DOMAIN(&screen
->base
) | NOUVEAU_BO_WR
);
942 for (i
= 0; i
< 5; ++i
) {
943 /* TIC and TSC entries for each unit (nve4+ only) */
944 /* auxiliary constants (6 user clip planes, base instance id) */
945 BEGIN_NVC0(push
, NVC0_3D(CB_SIZE
), 3);
946 PUSH_DATA (push
, 1024);
947 PUSH_DATAh(push
, screen
->uniform_bo
->offset
+ NVC0_CB_AUX_INFO(i
));
948 PUSH_DATA (push
, screen
->uniform_bo
->offset
+ NVC0_CB_AUX_INFO(i
));
949 BEGIN_NVC0(push
, NVC0_3D(CB_BIND(i
)), 1);
950 PUSH_DATA (push
, (15 << 4) | 1);
951 if (screen
->eng3d
->oclass
>= NVE4_3D_CLASS
) {
953 BEGIN_1IC0(push
, NVC0_3D(CB_POS
), 9);
954 PUSH_DATA (push
, NVC0_CB_AUX_UNK_INFO
);
955 for (j
= 0; j
< 8; ++j
)
958 BEGIN_NVC0(push
, NVC0_3D(TEX_LIMITS(i
)), 1);
959 PUSH_DATA (push
, 0x54);
962 BEGIN_NVC0(push
, NVC0_3D(LINKED_TSC
), 1);
965 /* return { 0.0, 0.0, 0.0, 0.0 } for out-of-bounds vtxbuf access */
966 BEGIN_NVC0(push
, NVC0_3D(CB_SIZE
), 3);
967 PUSH_DATA (push
, 256);
968 PUSH_DATAh(push
, screen
->uniform_bo
->offset
+ NVC0_CB_AUX_RUNOUT_INFO
);
969 PUSH_DATA (push
, screen
->uniform_bo
->offset
+ NVC0_CB_AUX_RUNOUT_INFO
);
970 BEGIN_1IC0(push
, NVC0_3D(CB_POS
), 5);
972 PUSH_DATAf(push
, 0.0f
);
973 PUSH_DATAf(push
, 0.0f
);
974 PUSH_DATAf(push
, 0.0f
);
975 PUSH_DATAf(push
, 0.0f
);
976 BEGIN_NVC0(push
, NVC0_3D(VERTEX_RUNOUT_ADDRESS_HIGH
), 2);
977 PUSH_DATAh(push
, screen
->uniform_bo
->offset
+ NVC0_CB_AUX_RUNOUT_INFO
);
978 PUSH_DATA (push
, screen
->uniform_bo
->offset
+ NVC0_CB_AUX_RUNOUT_INFO
);
980 if (screen
->base
.drm
->version
>= 0x01000101) {
981 ret
= nouveau_getparam(dev
, NOUVEAU_GETPARAM_GRAPH_UNITS
, &value
);
983 NOUVEAU_ERR("NOUVEAU_GETPARAM_GRAPH_UNITS failed.\n");
987 if (dev
->chipset
>= 0xe0 && dev
->chipset
< 0xf0)
988 value
= (8 << 8) | 4;
990 value
= (16 << 8) | 4;
992 screen
->gpc_count
= value
& 0x000000ff;
993 screen
->mp_count
= value
>> 8;
994 screen
->mp_count_compute
= screen
->mp_count
;
996 nvc0_screen_resize_tls_area(screen
, 128 * 16, 0, 0x200);
998 BEGIN_NVC0(push
, NVC0_3D(CODE_ADDRESS_HIGH
), 2);
999 PUSH_DATAh(push
, screen
->text
->offset
);
1000 PUSH_DATA (push
, screen
->text
->offset
);
1001 BEGIN_NVC0(push
, NVC0_3D(TEMP_ADDRESS_HIGH
), 4);
1002 PUSH_DATAh(push
, screen
->tls
->offset
);
1003 PUSH_DATA (push
, screen
->tls
->offset
);
1004 PUSH_DATA (push
, screen
->tls
->size
>> 32);
1005 PUSH_DATA (push
, screen
->tls
->size
);
1006 BEGIN_NVC0(push
, NVC0_3D(WARP_TEMP_ALLOC
), 1);
1007 PUSH_DATA (push
, 0);
1008 /* Reduce likelihood of collision with real buffers by placing the hole at
1009 * the top of the 4G area. This will have to be dealt with for real
1010 * eventually by blocking off that area from the VM.
1012 BEGIN_NVC0(push
, NVC0_3D(LOCAL_BASE
), 1);
1013 PUSH_DATA (push
, 0xff << 24);
1015 if (screen
->eng3d
->oclass
< GM107_3D_CLASS
) {
1016 ret
= nouveau_bo_new(dev
, NV_VRAM_DOMAIN(&screen
->base
), 1 << 17, 1 << 20, NULL
,
1017 &screen
->poly_cache
);
1021 BEGIN_NVC0(push
, NVC0_3D(VERTEX_QUARANTINE_ADDRESS_HIGH
), 3);
1022 PUSH_DATAh(push
, screen
->poly_cache
->offset
);
1023 PUSH_DATA (push
, screen
->poly_cache
->offset
);
1024 PUSH_DATA (push
, 3);
1027 ret
= nouveau_bo_new(dev
, NV_VRAM_DOMAIN(&screen
->base
), 1 << 17, 1 << 17, NULL
,
1032 BEGIN_NVC0(push
, NVC0_3D(TIC_ADDRESS_HIGH
), 3);
1033 PUSH_DATAh(push
, screen
->txc
->offset
);
1034 PUSH_DATA (push
, screen
->txc
->offset
);
1035 PUSH_DATA (push
, NVC0_TIC_MAX_ENTRIES
- 1);
1036 if (screen
->eng3d
->oclass
>= GM107_3D_CLASS
) {
1037 screen
->tic
.maxwell
= true;
1038 if (screen
->eng3d
->oclass
== GM107_3D_CLASS
) {
1039 screen
->tic
.maxwell
=
1040 debug_get_bool_option("NOUVEAU_MAXWELL_TIC", true);
1041 IMMED_NVC0(push
, SUBC_3D(0x0f10), screen
->tic
.maxwell
);
1045 BEGIN_NVC0(push
, NVC0_3D(TSC_ADDRESS_HIGH
), 3);
1046 PUSH_DATAh(push
, screen
->txc
->offset
+ 65536);
1047 PUSH_DATA (push
, screen
->txc
->offset
+ 65536);
1048 PUSH_DATA (push
, NVC0_TSC_MAX_ENTRIES
- 1);
1050 BEGIN_NVC0(push
, NVC0_3D(SCREEN_Y_CONTROL
), 1);
1051 PUSH_DATA (push
, 0);
1052 BEGIN_NVC0(push
, NVC0_3D(WINDOW_OFFSET_X
), 2);
1053 PUSH_DATA (push
, 0);
1054 PUSH_DATA (push
, 0);
1055 BEGIN_NVC0(push
, NVC0_3D(ZCULL_REGION
), 1); /* deactivate ZCULL */
1056 PUSH_DATA (push
, 0x3f);
1058 BEGIN_NVC0(push
, NVC0_3D(CLIP_RECTS_MODE
), 1);
1059 PUSH_DATA (push
, NVC0_3D_CLIP_RECTS_MODE_INSIDE_ANY
);
1060 BEGIN_NVC0(push
, NVC0_3D(CLIP_RECT_HORIZ(0)), 8 * 2);
1061 for (i
= 0; i
< 8 * 2; ++i
)
1063 BEGIN_NVC0(push
, NVC0_3D(CLIP_RECTS_EN
), 1);
1064 PUSH_DATA (push
, 0);
1065 BEGIN_NVC0(push
, NVC0_3D(CLIPID_ENABLE
), 1);
1066 PUSH_DATA (push
, 0);
1068 /* neither scissors, viewport nor stencil mask should affect clears */
1069 BEGIN_NVC0(push
, NVC0_3D(CLEAR_FLAGS
), 1);
1070 PUSH_DATA (push
, 0);
1072 BEGIN_NVC0(push
, NVC0_3D(VIEWPORT_TRANSFORM_EN
), 1);
1073 PUSH_DATA (push
, 1);
1074 for (i
= 0; i
< NVC0_MAX_VIEWPORTS
; i
++) {
1075 BEGIN_NVC0(push
, NVC0_3D(DEPTH_RANGE_NEAR(i
)), 2);
1076 PUSH_DATAf(push
, 0.0f
);
1077 PUSH_DATAf(push
, 1.0f
);
1079 BEGIN_NVC0(push
, NVC0_3D(VIEW_VOLUME_CLIP_CTRL
), 1);
1080 PUSH_DATA (push
, NVC0_3D_VIEW_VOLUME_CLIP_CTRL_UNK1_UNK1
);
1082 /* We use scissors instead of exact view volume clipping,
1083 * so they're always enabled.
1085 for (i
= 0; i
< NVC0_MAX_VIEWPORTS
; i
++) {
1086 BEGIN_NVC0(push
, NVC0_3D(SCISSOR_ENABLE(i
)), 3);
1087 PUSH_DATA (push
, 1);
1088 PUSH_DATA (push
, 8192 << 16);
1089 PUSH_DATA (push
, 8192 << 16);
1092 #define MK_MACRO(m, n) i = nvc0_graph_set_macro(screen, m, i, sizeof(n), n);
1095 MK_MACRO(NVC0_3D_MACRO_VERTEX_ARRAY_PER_INSTANCE
, mme9097_per_instance_bf
);
1096 MK_MACRO(NVC0_3D_MACRO_BLEND_ENABLES
, mme9097_blend_enables
);
1097 MK_MACRO(NVC0_3D_MACRO_VERTEX_ARRAY_SELECT
, mme9097_vertex_array_select
);
1098 MK_MACRO(NVC0_3D_MACRO_TEP_SELECT
, mme9097_tep_select
);
1099 MK_MACRO(NVC0_3D_MACRO_GP_SELECT
, mme9097_gp_select
);
1100 MK_MACRO(NVC0_3D_MACRO_POLYGON_MODE_FRONT
, mme9097_poly_mode_front
);
1101 MK_MACRO(NVC0_3D_MACRO_POLYGON_MODE_BACK
, mme9097_poly_mode_back
);
1102 MK_MACRO(NVC0_3D_MACRO_DRAW_ARRAYS_INDIRECT
, mme9097_draw_arrays_indirect
);
1103 MK_MACRO(NVC0_3D_MACRO_DRAW_ELEMENTS_INDIRECT
, mme9097_draw_elts_indirect
);
1104 MK_MACRO(NVC0_3D_MACRO_DRAW_ARRAYS_INDIRECT_COUNT
, mme9097_draw_arrays_indirect_count
);
1105 MK_MACRO(NVC0_3D_MACRO_DRAW_ELEMENTS_INDIRECT_COUNT
, mme9097_draw_elts_indirect_count
);
1106 MK_MACRO(NVC0_3D_MACRO_QUERY_BUFFER_WRITE
, mme9097_query_buffer_write
);
1107 MK_MACRO(NVC0_CP_MACRO_LAUNCH_GRID_INDIRECT
, mme90c0_launch_grid_indirect
);
1109 BEGIN_NVC0(push
, NVC0_3D(RASTERIZE_ENABLE
), 1);
1110 PUSH_DATA (push
, 1);
1111 BEGIN_NVC0(push
, NVC0_3D(RT_SEPARATE_FRAG_DATA
), 1);
1112 PUSH_DATA (push
, 1);
1113 BEGIN_NVC0(push
, NVC0_3D(MACRO_GP_SELECT
), 1);
1114 PUSH_DATA (push
, 0x40);
1115 BEGIN_NVC0(push
, NVC0_3D(LAYER
), 1);
1116 PUSH_DATA (push
, 0);
1117 BEGIN_NVC0(push
, NVC0_3D(MACRO_TEP_SELECT
), 1);
1118 PUSH_DATA (push
, 0x30);
1119 BEGIN_NVC0(push
, NVC0_3D(PATCH_VERTICES
), 1);
1120 PUSH_DATA (push
, 3);
1121 BEGIN_NVC0(push
, NVC0_3D(SP_SELECT(2)), 1);
1122 PUSH_DATA (push
, 0x20);
1123 BEGIN_NVC0(push
, NVC0_3D(SP_SELECT(0)), 1);
1124 PUSH_DATA (push
, 0x00);
1125 screen
->save_state
.patch_vertices
= 3;
1127 BEGIN_NVC0(push
, NVC0_3D(POINT_COORD_REPLACE
), 1);
1128 PUSH_DATA (push
, 0);
1129 BEGIN_NVC0(push
, NVC0_3D(POINT_RASTER_RULES
), 1);
1130 PUSH_DATA (push
, NVC0_3D_POINT_RASTER_RULES_OGL
);
1132 IMMED_NVC0(push
, NVC0_3D(EDGEFLAG
), 1);
1134 if (nvc0_screen_init_compute(screen
))
1139 screen
->tic
.entries
= CALLOC(4096, sizeof(void *));
1140 screen
->tsc
.entries
= screen
->tic
.entries
+ 2048;
1142 if (!nvc0_blitter_create(screen
))
1145 nouveau_fence_new(&screen
->base
, &screen
->base
.fence
.current
, false);
1147 return &screen
->base
;
1150 screen
->base
.base
.context_create
= NULL
;
1151 return &screen
->base
;
1155 nvc0_screen_tic_alloc(struct nvc0_screen
*screen
, void *entry
)
1157 int i
= screen
->tic
.next
;
1159 while (screen
->tic
.lock
[i
/ 32] & (1 << (i
% 32)))
1160 i
= (i
+ 1) & (NVC0_TIC_MAX_ENTRIES
- 1);
1162 screen
->tic
.next
= (i
+ 1) & (NVC0_TIC_MAX_ENTRIES
- 1);
1164 if (screen
->tic
.entries
[i
])
1165 nv50_tic_entry(screen
->tic
.entries
[i
])->id
= -1;
1167 screen
->tic
.entries
[i
] = entry
;
1172 nvc0_screen_tsc_alloc(struct nvc0_screen
*screen
, void *entry
)
1174 int i
= screen
->tsc
.next
;
1176 while (screen
->tsc
.lock
[i
/ 32] & (1 << (i
% 32)))
1177 i
= (i
+ 1) & (NVC0_TSC_MAX_ENTRIES
- 1);
1179 screen
->tsc
.next
= (i
+ 1) & (NVC0_TSC_MAX_ENTRIES
- 1);
1181 if (screen
->tsc
.entries
[i
])
1182 nv50_tsc_entry(screen
->tsc
.entries
[i
])->id
= -1;
1184 screen
->tsc
.entries
[i
] = entry
;